Abstract

Accelerating the learning curve of software maintainers working on systems with which they have little familiarity motivated this study. A working hypothesis was that automated methods are needed to provide a fast, rough grasp of a system, to enable practitioners not familiar with it, to commence maintenance with a level of confidence as if they had this familiarity. Expert maintainers were interviewed regarding their strategies and information needs to test this hypothesis. The overriding message is their need for a “starting point ” when analysing code. They also need standardised, reliable and communicable information about a system as an equivalent to knowledge available only to developers or experienced maintainers. These needs are addressed by the proposed “roughcut” approach to program comprehension. Work underway assesses the suitability of using data mining techniques on data derived from source code to provide high level models of a system and module interrelationships. 1. Background Program comprehension is a demanding task comprising up to 90 % of the total time spent on software maintenance [2], [21], [27], which in turn is the most expensive process in the lifetime of software [2]. This has been attributed to a plethora of problems reported in the literature, such as lack of up-to-date and precise documentation, inadequate communication, and unavailability of the original designers and programmers [11], [13]. Researchers have been trying to improve and accelerate the process of program comprehension in a number of ways. Because of the limited amount of information that a maintainer can assimilate at one time, the as-needed strategy suggests maintainers gain an understanding of the application while performing the change process [13]. Similarly partial comprehension has

Citations